Reframing AI Chatbots as Self-Service Tools

In a recent edition of the Supportive Weekly newsletter I wrote about why we might be better off treating AI chatbots as a tool for self-service, instead of as a full replacement for human support (which they just can’t completely achieve).

One reader, Ita, asked how we could practically achieve that reframing. In the latest episode of Supportive Shorts (the mini-episodes of The Supportive Podcast), I shared my answer:

The two elements are telling people what the tools are for, and using technical guardrails on the AI to align it with the same purposes. I share some specific examples in the podcast episode.

In case you can’t get enough of 1960s Australian television, here’s the full segment: 

'Instant money': ATM comes to Australia (1969) . You could only take out exactly $25, and your card had to be mailed back to you after each transaction. Handy.

ATMs didn’t perfectly replace bank tellers then, nor do they now. We had to learn how they were most helpfully used (or abused).

Now we’re all learning how AI tools work, the capabilities and limits of large language models, and how they can be best controlled to help deliver the form and quality of service we want.

If you’re looking for some LLM learning, here are two good options:

• Our own Principal Designer, Buzz, built an excellent introduction to AI course.

• DeepLearning.AI has plenty of clear and helpful resources.